2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved.
5 * This code is derived from software contributed to The DragonFly Project
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * Copyright (c) 1982, 1986, 1988, 1993
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
67 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
68 * $DragonFly: src/sys/netinet/ip_input.c,v 1.90 2008/08/23 04:12:23 sephe Exp $
73 #include "opt_bootp.h"
76 #include "opt_ipdivert.h"
77 #include "opt_ipfilter.h"
78 #include "opt_ipstealth.h"
79 #include "opt_ipsec.h"
81 #include <sys/param.h>
82 #include <sys/systm.h>
84 #include <sys/malloc.h>
85 #include <sys/mpipe.h>
86 #include <sys/domain.h>
87 #include <sys/protosw.h>
88 #include <sys/socket.h>
90 #include <sys/globaldata.h>
91 #include <sys/thread.h>
92 #include <sys/kernel.h>
93 #include <sys/syslog.h>
94 #include <sys/sysctl.h>
95 #include <sys/in_cksum.h>
97 #include <machine/stdarg.h>
100 #include <net/if_types.h>
101 #include <net/if_var.h>
102 #include <net/if_dl.h>
103 #include <net/pfil.h>
104 #include <net/route.h>
105 #include <net/netisr.h>
107 #include <netinet/in.h>
108 #include <netinet/in_systm.h>
109 #include <netinet/in_var.h>
110 #include <netinet/ip.h>
111 #include <netinet/in_pcb.h>
112 #include <netinet/ip_var.h>
113 #include <netinet/ip_icmp.h>
115 #include <sys/thread2.h>
116 #include <sys/msgport2.h>
117 #include <net/netmsg2.h>
119 #include <sys/socketvar.h>
121 #include <net/ipfw/ip_fw.h>
122 #include <net/dummynet/ip_dummynet.h>
125 #include <netinet6/ipsec.h>
126 #include <netproto/key/key.h>
130 #include <netproto/ipsec/ipsec.h>
131 #include <netproto/ipsec/key.h>
135 static int ip_rsvp_on
;
136 struct socket
*ip_rsvpd
;
138 int ipforwarding
= 0;
139 SYSCTL_INT(_net_inet_ip
, IPCTL_FORWARDING
, forwarding
, CTLFLAG_RW
,
140 &ipforwarding
, 0, "Enable IP forwarding between interfaces");
142 static int ipsendredirects
= 1; /* XXX */
143 SYSCTL_INT(_net_inet_ip
, IPCTL_SENDREDIRECTS
, redirect
, CTLFLAG_RW
,
144 &ipsendredirects
, 0, "Enable sending IP redirects");
146 int ip_defttl
= IPDEFTTL
;
147 SYSCTL_INT(_net_inet_ip
, IPCTL_DEFTTL
, ttl
, CTLFLAG_RW
,
148 &ip_defttl
, 0, "Maximum TTL on IP packets");
150 static int ip_dosourceroute
= 0;
151 SYSCTL_INT(_net_inet_ip
, IPCTL_SOURCEROUTE
, sourceroute
, CTLFLAG_RW
,
152 &ip_dosourceroute
, 0, "Enable forwarding source routed IP packets");
154 static int ip_acceptsourceroute
= 0;
155 SYSCTL_INT(_net_inet_ip
, IPCTL_ACCEPTSOURCEROUTE
, accept_sourceroute
,
156 CTLFLAG_RW
, &ip_acceptsourceroute
, 0,
157 "Enable accepting source routed IP packets");
159 static int ip_keepfaith
= 0;
160 SYSCTL_INT(_net_inet_ip
, IPCTL_KEEPFAITH
, keepfaith
, CTLFLAG_RW
,
162 "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
164 static int nipq
= 0; /* total # of reass queues */
166 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, maxfragpackets
, CTLFLAG_RW
,
168 "Maximum number of IPv4 fragment reassembly queue entries");
170 static int maxfragsperpacket
;
171 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, maxfragsperpacket
, CTLFLAG_RW
,
172 &maxfragsperpacket
, 0,
173 "Maximum number of IPv4 fragments allowed per packet");
175 static int ip_sendsourcequench
= 0;
176 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, sendsourcequench
, CTLFLAG_RW
,
177 &ip_sendsourcequench
, 0,
178 "Enable the transmission of source quench packets");
180 int ip_do_randomid
= 1;
181 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, random_id
, CTLFLAG_RW
,
183 "Assign random ip_id values");
185 * XXX - Setting ip_checkinterface mostly implements the receive side of
186 * the Strong ES model described in RFC 1122, but since the routing table
187 * and transmit implementation do not implement the Strong ES model,
188 * setting this to 1 results in an odd hybrid.
190 * XXX - ip_checkinterface currently must be disabled if you use ipnat
191 * to translate the destination address to another local interface.
193 * XXX - ip_checkinterface must be disabled if you add IP aliases
194 * to the loopback interface instead of the interface where the
195 * packets for those addresses are received.
197 static int ip_checkinterface
= 0;
198 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, check_interface
, CTLFLAG_RW
,
199 &ip_checkinterface
, 0, "Verify packet arrives on correct interface");
202 static int ipprintfs
= 0;
205 extern struct domain inetdomain
;
206 extern struct protosw inetsw
[];
207 u_char ip_protox
[IPPROTO_MAX
];
208 struct in_ifaddrhead in_ifaddrheads
[MAXCPU
]; /* first inet address */
209 struct in_ifaddrhashhead
*in_ifaddrhashtbls
[MAXCPU
];
210 /* inet addr hash table */
211 u_long in_ifaddrhmask
; /* mask for hash table */
213 struct ip_stats ipstats_percpu
[MAXCPU
];
216 sysctl_ipstats(SYSCTL_HANDLER_ARGS
)
220 for (cpu
= 0; cpu
< ncpus
; ++cpu
) {
221 if ((error
= SYSCTL_OUT(req
, &ipstats_percpu
[cpu
],
222 sizeof(struct ip_stats
))))
224 if ((error
= SYSCTL_IN(req
, &ipstats_percpu
[cpu
],
225 sizeof(struct ip_stats
))))
231 SYSCTL_PROC(_net_inet_ip
, IPCTL_STATS
, stats
, (CTLTYPE_OPAQUE
| CTLFLAG_RW
),
232 0, 0, sysctl_ipstats
, "S,ip_stats", "IP statistics");
234 SYSCTL_STRUCT(_net_inet_ip
, IPCTL_STATS
, stats
, CTLFLAG_RW
,
235 &ipstat
, ip_stats
, "IP statistics");
238 /* Packet reassembly stuff */
239 #define IPREASS_NHASH_LOG2 6
240 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
241 #define IPREASS_HMASK (IPREASS_NHASH - 1)
242 #define IPREASS_HASH(x,y) \
243 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
245 static struct ipq ipq
[IPREASS_NHASH
];
248 SYSCTL_INT(_net_inet_ip
, IPCTL_DEFMTU
, mtu
, CTLFLAG_RW
,
249 &ip_mtu
, 0, "Default MTU");
253 static int ipstealth
= 0;
254 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, stealth
, CTLFLAG_RW
, &ipstealth
, 0, "");
256 static const int ipstealth
= 0;
261 ip_fw_chk_t
*ip_fw_chk_ptr
;
262 ip_fw_dn_io_t
*ip_fw_dn_io_ptr
;
267 struct pfil_head inet_pfil_hook
;
270 * XXX this is ugly -- the following two global variables are
271 * used to store packet state while it travels through the stack.
272 * Note that the code even makes assumptions on the size and
273 * alignment of fields inside struct ip_srcrt so e.g. adding some
274 * fields will break the code. This needs to be fixed.
276 * We need to save the IP options in case a protocol wants to respond
277 * to an incoming packet over the same route if the packet got here
278 * using IP source routing. This allows connection establishment and
279 * maintenance when the remote end is on a network that is not known
282 static int ip_nhops
= 0;
284 static struct ip_srcrt
{
285 struct in_addr dst
; /* final destination */
286 char nop
; /* one NOP to align */
287 char srcopt
[IPOPT_OFFSET
+ 1]; /* OPTVAL, OLEN and OFFSET */
288 struct in_addr route
[MAX_IPOPTLEN
/sizeof(struct in_addr
)];
291 static MALLOC_DEFINE(M_IPQ
, "ipq", "IP Fragment Management");
292 static struct malloc_pipe ipq_mpipe
;
294 static void save_rte(u_char
*, struct in_addr
);
295 static int ip_dooptions(struct mbuf
*m
, int, struct sockaddr_in
*);
296 static void ip_freef(struct ipq
*);
297 static void ip_input_handler(struct netmsg
*);
298 static struct mbuf
*ip_reass(struct mbuf
*, u_int32_t
*);
301 * IP initialization: fill in IP protocol switch table.
302 * All protocols not implemented in kernel go to raw IP protocol handler.
314 * Make sure we can handle a reasonable number of fragments but
315 * cap it at 4000 (XXX).
317 mpipe_init(&ipq_mpipe
, M_IPQ
, sizeof(struct ipq
),
318 IFQ_MAXLEN
, 4000, 0, NULL
);
319 for (i
= 0; i
< ncpus
; ++i
) {
320 TAILQ_INIT(&in_ifaddrheads
[i
]);
321 in_ifaddrhashtbls
[i
] =
322 hashinit(INADDR_NHASH
, M_IFADDR
, &in_ifaddrhmask
);
324 pr
= pffindproto(PF_INET
, IPPROTO_RAW
, SOCK_RAW
);
327 for (i
= 0; i
< IPPROTO_MAX
; i
++)
328 ip_protox
[i
] = pr
- inetsw
;
329 for (pr
= inetdomain
.dom_protosw
;
330 pr
< inetdomain
.dom_protoswNPROTOSW
; pr
++)
331 if (pr
->pr_domain
->dom_family
== PF_INET
&&
332 pr
->pr_protocol
&& pr
->pr_protocol
!= IPPROTO_RAW
)
333 ip_protox
[pr
->pr_protocol
] = pr
- inetsw
;
335 inet_pfil_hook
.ph_type
= PFIL_TYPE_AF
;
336 inet_pfil_hook
.ph_af
= AF_INET
;
337 if ((i
= pfil_head_register(&inet_pfil_hook
)) != 0) {
338 kprintf("%s: WARNING: unable to register pfil hook, "
339 "error %d\n", __func__
, i
);
342 for (i
= 0; i
< IPREASS_NHASH
; i
++)
343 ipq
[i
].next
= ipq
[i
].prev
= &ipq
[i
];
345 maxnipq
= nmbclusters
/ 32;
346 maxfragsperpacket
= 16;
348 ip_id
= time_second
& 0xffff;
351 * Initialize IP statistics counters for each CPU.
355 for (cpu
= 0; cpu
< ncpus
; ++cpu
) {
356 bzero(&ipstats_percpu
[cpu
], sizeof(struct ip_stats
));
359 bzero(&ipstat
, sizeof(struct ip_stats
));
362 netisr_register(NETISR_IP
, ip_mport_in
, ip_input_handler
);
366 * XXX watch out this one. It is perhaps used as a cache for
367 * the most recently used route ? it is cleared in in_addroute()
368 * when a new route is successfully created.
370 struct route ipforward_rt
[MAXCPU
];
372 /* Do transport protocol processing. */
374 transport_processing_oncpu(struct mbuf
*m
, int hlen
, struct ip
*ip
)
377 * Switch out to protocol's input routine.
379 (*inetsw
[ip_protox
[ip
->ip_p
]].pr_input
)(m
, hlen
, ip
->ip_p
);
383 transport_processing_handler(netmsg_t netmsg
)
385 struct netmsg_packet
*pmsg
= (struct netmsg_packet
*)netmsg
;
389 ip
= mtod(pmsg
->nm_packet
, struct ip
*);
390 hlen
= pmsg
->nm_netmsg
.nm_lmsg
.u
.ms_result
;
392 transport_processing_oncpu(pmsg
->nm_packet
, hlen
, ip
);
393 /* netmsg was embedded in the mbuf, do not reply! */
397 ip_input_handler(struct netmsg
*msg0
)
399 struct mbuf
*m
= ((struct netmsg_packet
*)msg0
)->nm_packet
;
402 /* msg0 was embedded in the mbuf, do not reply! */
406 * IP input routine. Checksum and byte swap header. If fragmented
407 * try to reassemble. Process options. Pass to next level.
410 ip_input(struct mbuf
*m
)
413 struct in_ifaddr
*ia
= NULL
;
414 struct in_ifaddr_container
*iac
;
415 int i
, hlen
, checkif
;
417 struct in_addr pkt_dst
;
418 u_int32_t divert_info
= 0; /* packet divert/tee info */
419 struct ip_fw_args args
;
420 boolean_t using_srcrt
= FALSE
; /* forward (by PFIL_HOOKS) */
421 boolean_t needredispatch
= FALSE
;
422 struct in_addr odst
; /* original dst address(NAT) */
424 struct sockaddr_in
*next_hop
= NULL
;
426 struct tdb_ident
*tdbi
;
427 struct secpolicy
*sp
;
437 if (m
->m_pkthdr
.fw_flags
& IPFORWARD_MBUF_TAGGED
) {
439 mtag
= m_tag_find(m
, PACKET_TAG_IPFORWARD
, NULL
);
440 KKASSERT(mtag
!= NULL
);
441 next_hop
= m_tag_data(mtag
);
444 if (m
->m_pkthdr
.fw_flags
& DUMMYNET_MBUF_TAGGED
) {
445 /* Extract info from dummynet tag */
446 mtag
= m_tag_find(m
, PACKET_TAG_DUMMYNET
, NULL
);
447 KKASSERT(mtag
!= NULL
);
448 args
.rule
= ((struct dn_pkt
*)m_tag_data(mtag
))->dn_priv
;
449 KKASSERT(args
.rule
!= NULL
);
451 m_tag_delete(m
, mtag
);
452 m
->m_pkthdr
.fw_flags
&= ~DUMMYNET_MBUF_TAGGED
;
454 /* dummynet already filtered us */
455 ip
= mtod(m
, struct ip
*);
456 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
462 /* length checks already done in ip_demux() */
463 KASSERT(m
->m_len
>= sizeof(ip
), ("IP header not in one mbuf"));
465 ip
= mtod(m
, struct ip
*);
467 if (IP_VHL_V(ip
->ip_vhl
) != IPVERSION
) {
468 ipstat
.ips_badvers
++;
472 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
473 /* length checks already done in ip_demux() */
474 KASSERT(hlen
>= sizeof(struct ip
), ("IP header len too small"));
475 KASSERT(m
->m_len
>= hlen
, ("packet shorter than IP header length"));
477 /* 127/8 must not appear on wire - RFC1122 */
478 if ((ntohl(ip
->ip_dst
.s_addr
) >> IN_CLASSA_NSHIFT
) == IN_LOOPBACKNET
||
479 (ntohl(ip
->ip_src
.s_addr
) >> IN_CLASSA_NSHIFT
) == IN_LOOPBACKNET
) {
480 if (!(m
->m_pkthdr
.rcvif
->if_flags
& IFF_LOOPBACK
)) {
481 ipstat
.ips_badaddr
++;
486 if (m
->m_pkthdr
.csum_flags
& CSUM_IP_CHECKED
) {
487 sum
= !(m
->m_pkthdr
.csum_flags
& CSUM_IP_VALID
);
489 if (hlen
== sizeof(struct ip
)) {
490 sum
= in_cksum_hdr(ip
);
492 sum
= in_cksum(m
, hlen
);
501 if (altq_input
!= NULL
&& (*altq_input
)(m
, AF_INET
) == 0) {
502 /* packet is dropped by traffic conditioner */
507 * Convert fields to host representation.
509 ip
->ip_len
= ntohs(ip
->ip_len
);
510 if (ip
->ip_len
< hlen
) {
514 ip
->ip_off
= ntohs(ip
->ip_off
);
517 * Check that the amount of data in the buffers
518 * is as at least much as the IP header would have us expect.
519 * Trim mbufs if longer than we expect.
520 * Drop packet if shorter than we expect.
522 if (m
->m_pkthdr
.len
< ip
->ip_len
) {
523 ipstat
.ips_tooshort
++;
526 if (m
->m_pkthdr
.len
> ip
->ip_len
) {
527 if (m
->m_len
== m
->m_pkthdr
.len
) {
528 m
->m_len
= ip
->ip_len
;
529 m
->m_pkthdr
.len
= ip
->ip_len
;
531 m_adj(m
, ip
->ip_len
- m
->m_pkthdr
.len
);
533 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
535 * Bypass packet filtering for packets from a tunnel (gif).
537 if (ipsec_gethist(m
, NULL
))
543 * Right now when no processing on packet has done
544 * and it is still fresh out of network we do our black
546 * - Firewall: deny/allow/divert
547 * - Xlate: translate packet's addr/port (NAT).
548 * - Pipe: pass pkt through dummynet.
549 * - Wrap: fake packet's addr/port <unimpl.>
550 * - Encapsulate: put it in another IP and send out. <unimp.>
556 * Run through list of hooks for input packets.
558 * NB: Beware of the destination address changing (e.g.
559 * by NAT rewriting). When this happens, tell
560 * ip_forward to do the right thing.
562 if (pfil_has_hooks(&inet_pfil_hook
)) {
564 if (pfil_run_hooks(&inet_pfil_hook
, &m
,
565 m
->m_pkthdr
.rcvif
, PFIL_IN
)) {
568 if (m
== NULL
) /* consumed by filter */
570 ip
= mtod(m
, struct ip
*);
571 using_srcrt
= (odst
.s_addr
!= ip
->ip_dst
.s_addr
);
574 if (fw_enable
&& IPFW_LOADED
) {
576 * If we've been forwarded from the output side, then
577 * skip the firewall a second time
579 if (next_hop
!= NULL
)
583 i
= ip_fw_chk_ptr(&args
);
586 if ((i
& IP_FW_PORT_DENY_FLAG
) || m
== NULL
) { /* drop */
591 ip
= mtod(m
, struct ip
*); /* just in case m changed */
593 if (m
->m_pkthdr
.fw_flags
& IPFORWARD_MBUF_TAGGED
) {
594 mtag
= m_tag_find(m
, PACKET_TAG_IPFORWARD
, NULL
);
595 KKASSERT(mtag
!= NULL
);
596 next_hop
= m_tag_data(mtag
);
599 if (i
== 0 && next_hop
== NULL
) /* common case */
601 if (i
& IP_FW_PORT_DYNT_FLAG
) {
602 /* Send packet to the appropriate pipe */
603 ip_fw_dn_io_ptr(m
, i
&0xffff, DN_TO_IP_IN
, &args
);
607 if (i
!= 0 && !(i
& IP_FW_PORT_DYNT_FLAG
)) {
608 /* Divert or tee packet */
613 if (i
== 0 && next_hop
!= NULL
)
616 * if we get here, the packet must be dropped
624 * Process options and, if not destined for us,
625 * ship it on. ip_dooptions returns 1 when an
626 * error was detected (causing an icmp message
627 * to be sent and the original packet to be freed).
629 ip_nhops
= 0; /* for source routed packets */
630 if (hlen
> sizeof(struct ip
) && ip_dooptions(m
, 0, next_hop
))
633 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
634 * matter if it is destined to another node, or whether it is
635 * a multicast one, RSVP wants it! and prevents it from being forwarded
636 * anywhere else. Also checks if the rsvp daemon is running before
637 * grabbing the packet.
639 if (rsvp_on
&& ip
->ip_p
== IPPROTO_RSVP
)
643 * Check our list of addresses, to see if the packet is for us.
644 * If we don't have any addresses, assume any unicast packet
645 * we receive might be for us (and let the upper layers deal
648 if (TAILQ_EMPTY(&in_ifaddrheads
[mycpuid
]) &&
649 !(m
->m_flags
& (M_MCAST
| M_BCAST
)))
653 * Cache the destination address of the packet; this may be
654 * changed by use of 'ipfw fwd'.
656 pkt_dst
= next_hop
? next_hop
->sin_addr
: ip
->ip_dst
;
659 * Enable a consistency check between the destination address
660 * and the arrival interface for a unicast packet (the RFC 1122
661 * strong ES model) if IP forwarding is disabled and the packet
662 * is not locally generated and the packet is not subject to
665 * XXX - Checking also should be disabled if the destination
666 * address is ipnat'ed to a different interface.
668 * XXX - Checking is incompatible with IP aliases added
669 * to the loopback interface instead of the interface where
670 * the packets are received.
672 checkif
= ip_checkinterface
&&
674 m
->m_pkthdr
.rcvif
!= NULL
&&
675 !(m
->m_pkthdr
.rcvif
->if_flags
& IFF_LOOPBACK
) &&
679 * Check for exact addresses in the hash bucket.
681 LIST_FOREACH(iac
, INADDR_HASH(pkt_dst
.s_addr
), ia_hash
) {
685 * If the address matches, verify that the packet
686 * arrived via the correct interface if checking is
689 if (IA_SIN(ia
)->sin_addr
.s_addr
== pkt_dst
.s_addr
&&
690 (!checkif
|| ia
->ia_ifp
== m
->m_pkthdr
.rcvif
))
696 * Check for broadcast addresses.
698 * Only accept broadcast packets that arrive via the matching
699 * interface. Reception of forwarded directed broadcasts would
700 * be handled via ip_forward() and ether_output() with the loopback
701 * into the stack for SIMPLEX interfaces handled by ether_output().
703 if (m
->m_pkthdr
.rcvif
->if_flags
& IFF_BROADCAST
) {
704 struct ifaddr_container
*ifac
;
706 TAILQ_FOREACH(ifac
, &m
->m_pkthdr
.rcvif
->if_addrheads
[mycpuid
],
708 struct ifaddr
*ifa
= ifac
->ifa
;
710 if (ifa
->ifa_addr
== NULL
) /* shutdown/startup race */
712 if (ifa
->ifa_addr
->sa_family
!= AF_INET
)
715 if (satosin(&ia
->ia_broadaddr
)->sin_addr
.s_addr
==
718 if (ia
->ia_netbroadcast
.s_addr
== pkt_dst
.s_addr
)
721 if (IA_SIN(ia
)->sin_addr
.s_addr
== INADDR_ANY
)
726 if (IN_MULTICAST(ntohl(ip
->ip_dst
.s_addr
))) {
727 struct in_multi
*inm
;
729 if (ip_mrouter
!= NULL
) {
731 * If we are acting as a multicast router, all
732 * incoming multicast packets are passed to the
733 * kernel-level multicast forwarding function.
734 * The packet is returned (relatively) intact; if
735 * ip_mforward() returns a non-zero value, the packet
736 * must be discarded, else it may be accepted below.
738 if (ip_mforward
!= NULL
&&
739 ip_mforward(ip
, m
->m_pkthdr
.rcvif
, m
, NULL
) != 0) {
740 ipstat
.ips_cantforward
++;
746 * The process-level routing daemon needs to receive
747 * all multicast IGMP packets, whether or not this
748 * host belongs to their destination groups.
750 if (ip
->ip_p
== IPPROTO_IGMP
)
752 ipstat
.ips_forward
++;
755 * See if we belong to the destination multicast group on the
758 IN_LOOKUP_MULTI(ip
->ip_dst
, m
->m_pkthdr
.rcvif
, inm
);
760 ipstat
.ips_notmember
++;
766 if (ip
->ip_dst
.s_addr
== INADDR_BROADCAST
)
768 if (ip
->ip_dst
.s_addr
== INADDR_ANY
)
772 * FAITH(Firewall Aided Internet Translator)
774 if (m
->m_pkthdr
.rcvif
&& m
->m_pkthdr
.rcvif
->if_type
== IFT_FAITH
) {
776 if (ip
->ip_p
== IPPROTO_TCP
|| ip
->ip_p
== IPPROTO_ICMP
)
784 * Not for us; forward if possible and desirable.
787 ipstat
.ips_cantforward
++;
792 * Enforce inbound IPsec SPD.
794 if (ipsec4_in_reject(m
, NULL
)) {
795 ipsecstat
.in_polvio
++;
800 mtag
= m_tag_find(m
, PACKET_TAG_IPSEC_IN_DONE
, NULL
);
803 tdbi
= (struct tdb_ident
*)m_tag_data(mtag
);
804 sp
= ipsec_getpolicy(tdbi
, IPSEC_DIR_INBOUND
);
806 sp
= ipsec_getpolicybyaddr(m
, IPSEC_DIR_INBOUND
,
807 IP_FORWARDING
, &error
);
809 if (sp
== NULL
) { /* NB: can happen if error */
811 /*XXX error stat???*/
812 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
817 * Check security policy against packet attributes.
819 error
= ipsec_in_reject(sp
, m
);
823 ipstat
.ips_cantforward
++;
827 ip_forward(m
, using_srcrt
, next_hop
);
834 * IPSTEALTH: Process non-routing options only
835 * if the packet is destined for us.
838 hlen
> sizeof(struct ip
) &&
839 ip_dooptions(m
, 1, next_hop
))
842 /* Count the packet in the ip address stats */
844 ia
->ia_ifa
.if_ipackets
++;
845 ia
->ia_ifa
.if_ibytes
+= m
->m_pkthdr
.len
;
849 * If offset or IP_MF are set, must reassemble.
850 * Otherwise, nothing need be done.
851 * (We could look in the reassembly queue to see
852 * if the packet was previously fragmented,
853 * but it's not worth the time; just let them time out.)
855 if (ip
->ip_off
& (IP_MF
| IP_OFFMASK
)) {
857 * Attempt reassembly; if it succeeds, proceed.
858 * ip_reass() will return a different mbuf, and update
859 * the divert info in divert_info.
861 m
= ip_reass(m
, &divert_info
);
865 needredispatch
= TRUE
;
866 ip
= mtod(m
, struct ip
*);
867 /* Get the header length of the reassembled packet */
868 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
870 /* Restore original checksum before diverting packet */
871 if (divert_info
!= 0) {
873 ip
->ip_len
= htons(ip
->ip_len
);
874 ip
->ip_off
= htons(ip
->ip_off
);
876 if (hlen
== sizeof(struct ip
))
877 ip
->ip_sum
= in_cksum_hdr(ip
);
879 ip
->ip_sum
= in_cksum(m
, hlen
);
880 ip
->ip_off
= ntohs(ip
->ip_off
);
881 ip
->ip_len
= ntohs(ip
->ip_len
);
891 * Divert or tee packet to the divert protocol if required.
893 if (divert_info
!= 0) {
894 struct mbuf
*clone
= NULL
;
896 /* Clone packet if we're doing a 'tee' */
897 if ((divert_info
& IP_FW_PORT_TEE_FLAG
) != 0)
898 clone
= m_dup(m
, MB_DONTWAIT
);
900 /* Restore packet header fields to original values */
902 ip
->ip_len
= htons(ip
->ip_len
);
903 ip
->ip_off
= htons(ip
->ip_off
);
905 /* Deliver packet to divert input routine */
906 divert_packet(m
, 1, divert_info
& 0xffff);
907 ipstat
.ips_delivered
++;
909 /* If 'tee', continue with original packet */
913 ip
= mtod(m
, struct ip
*);
916 * Jump backwards to complete processing of the
917 * packet. But first clear divert_info to avoid
918 * entering this block again.
919 * We do not need to clear args.divert_rule as
920 * it will not be used.
922 * XXX Better safe than sorry, remove the DIVERT tag.
924 mtag
= m_tag_find(m
, PACKET_TAG_IPFW_DIVERT
, NULL
);
926 m_tag_delete(m
, mtag
);
935 * enforce IPsec policy checking if we are seeing last header.
936 * note that we do not visit this with protocols with pcb layer
937 * code - like udp/tcp/raw ip.
939 if ((inetsw
[ip_protox
[ip
->ip_p
]].pr_flags
& PR_LASTHDR
) &&
940 ipsec4_in_reject(m
, NULL
)) {
941 ipsecstat
.in_polvio
++;
947 * enforce IPsec policy checking if we are seeing last header.
948 * note that we do not visit this with protocols with pcb layer
949 * code - like udp/tcp/raw ip.
951 if (inetsw
[ip_protox
[ip
->ip_p
]].pr_flags
& PR_LASTHDR
) {
953 * Check if the packet has already had IPsec processing
954 * done. If so, then just pass it along. This tag gets
955 * set during AH, ESP, etc. input handling, before the
956 * packet is returned to the ip input queue for delivery.
958 mtag
= m_tag_find(m
, PACKET_TAG_IPSEC_IN_DONE
, NULL
);
961 tdbi
= (struct tdb_ident
*)m_tag_data(mtag
);
962 sp
= ipsec_getpolicy(tdbi
, IPSEC_DIR_INBOUND
);
964 sp
= ipsec_getpolicybyaddr(m
, IPSEC_DIR_INBOUND
,
965 IP_FORWARDING
, &error
);
969 * Check security policy against packet attributes.
971 error
= ipsec_in_reject(sp
, m
);
974 /* XXX error stat??? */
976 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
983 #endif /* FAST_IPSEC */
985 ipstat
.ips_delivered
++;
986 if (needredispatch
) {
987 struct netmsg_packet
*pmsg
;
990 ip
->ip_off
= htons(ip
->ip_off
);
991 ip
->ip_len
= htons(ip
->ip_len
);
992 port
= ip_mport_in(&m
);
996 pmsg
= &m
->m_hdr
.mh_netmsg
;
997 netmsg_init(&pmsg
->nm_netmsg
, &netisr_apanic_rport
, 0,
998 transport_processing_handler
);
1000 pmsg
->nm_netmsg
.nm_lmsg
.u
.ms_result
= hlen
;
1002 ip
= mtod(m
, struct ip
*);
1003 ip
->ip_len
= ntohs(ip
->ip_len
);
1004 ip
->ip_off
= ntohs(ip
->ip_off
);
1005 lwkt_sendmsg(port
, &pmsg
->nm_netmsg
.nm_lmsg
);
1007 transport_processing_oncpu(m
, hlen
, ip
);
1016 * Take incoming datagram fragment and try to reassemble it into
1017 * whole datagram. If a chain for reassembly of this datagram already
1018 * exists, then it is given as fp; otherwise have to make a chain.
1020 * When IPDIVERT enabled, keep additional state with each packet that
1021 * tells us if we need to divert or tee the packet we're building.
1022 * In particular, *divinfo includes the port and TEE flag.
1025 static struct mbuf
*
1026 ip_reass(struct mbuf
*m
, u_int32_t
*divinfo
)
1028 struct ip
*ip
= mtod(m
, struct ip
*);
1029 struct mbuf
*p
= NULL
, *q
, *nq
;
1031 struct ipq
*fp
= NULL
;
1032 int hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
1039 /* If maxnipq is 0, never accept fragments. */
1041 ipstat
.ips_fragments
++;
1042 ipstat
.ips_fragdropped
++;
1047 sum
= IPREASS_HASH(ip
->ip_src
.s_addr
, ip
->ip_id
);
1049 * Look for queue of fragments of this datagram.
1051 for (fp
= ipq
[sum
].next
; fp
!= &ipq
[sum
]; fp
= fp
->next
)
1052 if (ip
->ip_id
== fp
->ipq_id
&&
1053 ip
->ip_src
.s_addr
== fp
->ipq_src
.s_addr
&&
1054 ip
->ip_dst
.s_addr
== fp
->ipq_dst
.s_addr
&&
1055 ip
->ip_p
== fp
->ipq_p
)
1061 * Enforce upper bound on number of fragmented packets
1062 * for which we attempt reassembly;
1063 * If maxnipq is -1, accept all fragments without limitation.
1065 if (nipq
> maxnipq
&& maxnipq
> 0) {
1067 * drop something from the tail of the current queue
1068 * before proceeding further
1070 if (ipq
[sum
].prev
== &ipq
[sum
]) { /* gak */
1071 for (i
= 0; i
< IPREASS_NHASH
; i
++) {
1072 if (ipq
[i
].prev
!= &ipq
[i
]) {
1073 ipstat
.ips_fragtimeout
+=
1074 ipq
[i
].prev
->ipq_nfrags
;
1075 ip_freef(ipq
[i
].prev
);
1080 ipstat
.ips_fragtimeout
+=
1081 ipq
[sum
].prev
->ipq_nfrags
;
1082 ip_freef(ipq
[sum
].prev
);
1087 * Adjust ip_len to not reflect header,
1088 * convert offset of this to bytes.
1091 if (ip
->ip_off
& IP_MF
) {
1093 * Make sure that fragments have a data length
1094 * that's a non-zero multiple of 8 bytes.
1096 if (ip
->ip_len
== 0 || (ip
->ip_len
& 0x7) != 0) {
1097 ipstat
.ips_toosmall
++; /* XXX */
1101 m
->m_flags
|= M_FRAG
;
1103 m
->m_flags
&= ~M_FRAG
;
1106 ipstat
.ips_fragments
++;
1107 m
->m_pkthdr
.header
= ip
;
1110 * If the hardware has not done csum over this fragment
1111 * then csum_data is not valid at all.
1113 if ((m
->m_pkthdr
.csum_flags
& (CSUM_FRAG_NOT_CHECKED
| CSUM_DATA_VALID
))
1114 == (CSUM_FRAG_NOT_CHECKED
| CSUM_DATA_VALID
)) {
1115 m
->m_pkthdr
.csum_data
= 0;
1116 m
->m_pkthdr
.csum_flags
&= ~(CSUM_DATA_VALID
| CSUM_PSEUDO_HDR
);
1120 * Presence of header sizes in mbufs
1121 * would confuse code below.
1127 * If first fragment to arrive, create a reassembly queue.
1130 if ((fp
= mpipe_alloc_nowait(&ipq_mpipe
)) == NULL
)
1132 insque(fp
, &ipq
[sum
]);
1135 fp
->ipq_ttl
= IPFRAGTTL
;
1136 fp
->ipq_p
= ip
->ip_p
;
1137 fp
->ipq_id
= ip
->ip_id
;
1138 fp
->ipq_src
= ip
->ip_src
;
1139 fp
->ipq_dst
= ip
->ip_dst
;
1141 m
->m_nextpkt
= NULL
;
1143 fp
->ipq_div_info
= 0;
1150 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1153 * Find a segment which begins after this one does.
1155 for (p
= NULL
, q
= fp
->ipq_frags
; q
; p
= q
, q
= q
->m_nextpkt
)
1156 if (GETIP(q
)->ip_off
> ip
->ip_off
)
1160 * If there is a preceding segment, it may provide some of
1161 * our data already. If so, drop the data from the incoming
1162 * segment. If it provides all of our data, drop us, otherwise
1163 * stick new segment in the proper place.
1165 * If some of the data is dropped from the the preceding
1166 * segment, then it's checksum is invalidated.
1169 i
= GETIP(p
)->ip_off
+ GETIP(p
)->ip_len
- ip
->ip_off
;
1171 if (i
>= ip
->ip_len
)
1174 m
->m_pkthdr
.csum_flags
= 0;
1178 m
->m_nextpkt
= p
->m_nextpkt
;
1181 m
->m_nextpkt
= fp
->ipq_frags
;
1186 * While we overlap succeeding segments trim them or,
1187 * if they are completely covered, dequeue them.
1189 for (; q
!= NULL
&& ip
->ip_off
+ ip
->ip_len
> GETIP(q
)->ip_off
;
1191 i
= (ip
->ip_off
+ ip
->ip_len
) - GETIP(q
)->ip_off
;
1192 if (i
< GETIP(q
)->ip_len
) {
1193 GETIP(q
)->ip_len
-= i
;
1194 GETIP(q
)->ip_off
+= i
;
1196 q
->m_pkthdr
.csum_flags
= 0;
1201 ipstat
.ips_fragdropped
++;
1203 q
->m_nextpkt
= NULL
;
1211 * Transfer firewall instructions to the fragment structure.
1212 * Only trust info in the fragment at offset 0.
1214 if (ip
->ip_off
== 0) {
1215 fp
->ipq_div_info
= *divinfo
;
1217 mtag
= m_tag_find(m
, PACKET_TAG_IPFW_DIVERT
, NULL
);
1219 m_tag_delete(m
, mtag
);
1225 * Check for complete reassembly and perform frag per packet
1228 * Frag limiting is performed here so that the nth frag has
1229 * a chance to complete the packet before we drop the packet.
1230 * As a result, n+1 frags are actually allowed per packet, but
1231 * only n will ever be stored. (n = maxfragsperpacket.)
1235 for (p
= NULL
, q
= fp
->ipq_frags
; q
; p
= q
, q
= q
->m_nextpkt
) {
1236 if (GETIP(q
)->ip_off
!= next
) {
1237 if (fp
->ipq_nfrags
> maxfragsperpacket
) {
1238 ipstat
.ips_fragdropped
+= fp
->ipq_nfrags
;
1243 next
+= GETIP(q
)->ip_len
;
1245 /* Make sure the last packet didn't have the IP_MF flag */
1246 if (p
->m_flags
& M_FRAG
) {
1247 if (fp
->ipq_nfrags
> maxfragsperpacket
) {
1248 ipstat
.ips_fragdropped
+= fp
->ipq_nfrags
;
1255 * Reassembly is complete. Make sure the packet is a sane size.
1259 if (next
+ (IP_VHL_HL(ip
->ip_vhl
) << 2) > IP_MAXPACKET
) {
1260 ipstat
.ips_toolong
++;
1261 ipstat
.ips_fragdropped
+= fp
->ipq_nfrags
;
1267 * Concatenate fragments.
1274 q
->m_nextpkt
= NULL
;
1275 for (q
= nq
; q
!= NULL
; q
= nq
) {
1277 q
->m_nextpkt
= NULL
;
1278 m
->m_pkthdr
.csum_flags
&= q
->m_pkthdr
.csum_flags
;
1279 m
->m_pkthdr
.csum_data
+= q
->m_pkthdr
.csum_data
;
1284 * Clean up the 1's complement checksum. Carry over 16 bits must
1285 * be added back. This assumes no more then 65535 packet fragments
1286 * were reassembled. A second carry can also occur (but not a third).
1288 m
->m_pkthdr
.csum_data
= (m
->m_pkthdr
.csum_data
& 0xffff) +
1289 (m
->m_pkthdr
.csum_data
>> 16);
1290 if (m
->m_pkthdr
.csum_data
> 0xFFFF)
1291 m
->m_pkthdr
.csum_data
-= 0xFFFF;
1296 * Extract firewall instructions from the fragment structure.
1298 *divinfo
= fp
->ipq_div_info
;
1302 * Create header for new ip packet by
1303 * modifying header of first packet;
1304 * dequeue and discard fragment reassembly header.
1305 * Make header visible.
1308 ip
->ip_src
= fp
->ipq_src
;
1309 ip
->ip_dst
= fp
->ipq_dst
;
1312 mpipe_free(&ipq_mpipe
, fp
);
1313 m
->m_len
+= (IP_VHL_HL(ip
->ip_vhl
) << 2);
1314 m
->m_data
-= (IP_VHL_HL(ip
->ip_vhl
) << 2);
1315 /* some debugging cruft by sklower, below, will go away soon */
1316 if (m
->m_flags
& M_PKTHDR
) { /* XXX this should be done elsewhere */
1319 for (n
= m
; n
; n
= n
->m_next
)
1321 m
->m_pkthdr
.len
= plen
;
1324 ipstat
.ips_reassembled
++;
1331 ipstat
.ips_fragdropped
++;
1341 * Free a fragment reassembly header and all
1342 * associated datagrams.
1345 ip_freef(struct ipq
*fp
)
1349 while (fp
->ipq_frags
) {
1351 fp
->ipq_frags
= q
->m_nextpkt
;
1352 q
->m_nextpkt
= NULL
;
1356 mpipe_free(&ipq_mpipe
, fp
);
1361 * IP timer processing;
1362 * if a timer expires on a reassembly
1363 * queue, discard it.
1372 for (i
= 0; i
< IPREASS_NHASH
; i
++) {
1376 while (fp
!= &ipq
[i
]) {
1379 if (fp
->prev
->ipq_ttl
== 0) {
1380 ipstat
.ips_fragtimeout
+= fp
->prev
->ipq_nfrags
;
1386 * If we are over the maximum number of fragments
1387 * (due to the limit being lowered), drain off
1388 * enough to get down to the new limit.
1390 if (maxnipq
>= 0 && nipq
> maxnipq
) {
1391 for (i
= 0; i
< IPREASS_NHASH
; i
++) {
1392 while (nipq
> maxnipq
&&
1393 (ipq
[i
].next
!= &ipq
[i
])) {
1394 ipstat
.ips_fragdropped
+=
1395 ipq
[i
].next
->ipq_nfrags
;
1396 ip_freef(ipq
[i
].next
);
1405 * Drain off all datagram fragments.
1412 for (i
= 0; i
< IPREASS_NHASH
; i
++) {
1413 while (ipq
[i
].next
!= &ipq
[i
]) {
1414 ipstat
.ips_fragdropped
+= ipq
[i
].next
->ipq_nfrags
;
1415 ip_freef(ipq
[i
].next
);
1422 * Do option processing on a datagram,
1423 * possibly discarding it if bad options are encountered,
1424 * or forwarding it if source-routed.
1425 * The pass argument is used when operating in the IPSTEALTH
1426 * mode to tell what options to process:
1427 * [LS]SRR (pass 0) or the others (pass 1).
1428 * The reason for as many as two passes is that when doing IPSTEALTH,
1429 * non-routing options should be processed only if the packet is for us.
1430 * Returns 1 if packet has been forwarded/freed,
1431 * 0 if the packet should be processed further.
1434 ip_dooptions(struct mbuf
*m
, int pass
, struct sockaddr_in
*next_hop
)
1436 struct sockaddr_in ipaddr
= { sizeof ipaddr
, AF_INET
};
1437 struct ip
*ip
= mtod(m
, struct ip
*);
1439 struct in_ifaddr
*ia
;
1440 int opt
, optlen
, cnt
, off
, code
, type
= ICMP_PARAMPROB
;
1441 boolean_t forward
= FALSE
;
1442 struct in_addr
*sin
, dst
;
1446 cp
= (u_char
*)(ip
+ 1);
1447 cnt
= (IP_VHL_HL(ip
->ip_vhl
) << 2) - sizeof(struct ip
);
1448 for (; cnt
> 0; cnt
-= optlen
, cp
+= optlen
) {
1449 opt
= cp
[IPOPT_OPTVAL
];
1450 if (opt
== IPOPT_EOL
)
1452 if (opt
== IPOPT_NOP
)
1455 if (cnt
< IPOPT_OLEN
+ sizeof(*cp
)) {
1456 code
= &cp
[IPOPT_OLEN
] - (u_char
*)ip
;
1459 optlen
= cp
[IPOPT_OLEN
];
1460 if (optlen
< IPOPT_OLEN
+ sizeof(*cp
) || optlen
> cnt
) {
1461 code
= &cp
[IPOPT_OLEN
] - (u_char
*)ip
;
1471 * Source routing with record.
1472 * Find interface with current destination address.
1473 * If none on this machine then drop if strictly routed,
1474 * or do nothing if loosely routed.
1475 * Record interface address and bring up next address
1476 * component. If strictly routed make sure next
1477 * address is on directly accessible net.
1481 if (ipstealth
&& pass
> 0)
1483 if (optlen
< IPOPT_OFFSET
+ sizeof(*cp
)) {
1484 code
= &cp
[IPOPT_OLEN
] - (u_char
*)ip
;
1487 if ((off
= cp
[IPOPT_OFFSET
]) < IPOPT_MINOFF
) {
1488 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1491 ipaddr
.sin_addr
= ip
->ip_dst
;
1492 ia
= (struct in_ifaddr
*)
1493 ifa_ifwithaddr((struct sockaddr
*)&ipaddr
);
1495 if (opt
== IPOPT_SSRR
) {
1496 type
= ICMP_UNREACH
;
1497 code
= ICMP_UNREACH_SRCFAIL
;
1500 if (!ip_dosourceroute
)
1501 goto nosourcerouting
;
1503 * Loose routing, and not at next destination
1504 * yet; nothing to do except forward.
1508 off
--; /* 0 origin */
1509 if (off
> optlen
- (int)sizeof(struct in_addr
)) {
1511 * End of source route. Should be for us.
1513 if (!ip_acceptsourceroute
)
1514 goto nosourcerouting
;
1515 save_rte(cp
, ip
->ip_src
);
1520 if (!ip_dosourceroute
) {
1522 char buf
[sizeof "aaa.bbb.ccc.ddd"];
1525 * Acting as a router, so generate ICMP
1528 strcpy(buf
, inet_ntoa(ip
->ip_dst
));
1530 "attempted source route from %s to %s\n",
1531 inet_ntoa(ip
->ip_src
), buf
);
1532 type
= ICMP_UNREACH
;
1533 code
= ICMP_UNREACH_SRCFAIL
;
1537 * Not acting as a router,
1541 ipstat
.ips_cantforward
++;
1548 * locate outgoing interface
1550 memcpy(&ipaddr
.sin_addr
, cp
+ off
,
1551 sizeof ipaddr
.sin_addr
);
1553 if (opt
== IPOPT_SSRR
) {
1554 #define INA struct in_ifaddr *
1555 #define SA struct sockaddr *
1556 if ((ia
= (INA
)ifa_ifwithdstaddr((SA
)&ipaddr
))
1558 ia
= (INA
)ifa_ifwithnet((SA
)&ipaddr
);
1560 ia
= ip_rtaddr(ipaddr
.sin_addr
,
1561 &ipforward_rt
[mycpuid
]);
1563 type
= ICMP_UNREACH
;
1564 code
= ICMP_UNREACH_SRCFAIL
;
1567 ip
->ip_dst
= ipaddr
.sin_addr
;
1568 memcpy(cp
+ off
, &IA_SIN(ia
)->sin_addr
,
1569 sizeof(struct in_addr
));
1570 cp
[IPOPT_OFFSET
] += sizeof(struct in_addr
);
1572 * Let ip_intr's mcast routing check handle mcast pkts
1574 forward
= !IN_MULTICAST(ntohl(ip
->ip_dst
.s_addr
));
1578 if (ipstealth
&& pass
== 0)
1580 if (optlen
< IPOPT_OFFSET
+ sizeof(*cp
)) {
1581 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1584 if ((off
= cp
[IPOPT_OFFSET
]) < IPOPT_MINOFF
) {
1585 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1589 * If no space remains, ignore.
1591 off
--; /* 0 origin */
1592 if (off
> optlen
- (int)sizeof(struct in_addr
))
1594 memcpy(&ipaddr
.sin_addr
, &ip
->ip_dst
,
1595 sizeof ipaddr
.sin_addr
);
1597 * locate outgoing interface; if we're the destination,
1598 * use the incoming interface (should be same).
1600 if ((ia
= (INA
)ifa_ifwithaddr((SA
)&ipaddr
)) == NULL
&&
1601 (ia
= ip_rtaddr(ipaddr
.sin_addr
,
1602 &ipforward_rt
[mycpuid
]))
1604 type
= ICMP_UNREACH
;
1605 code
= ICMP_UNREACH_HOST
;
1608 memcpy(cp
+ off
, &IA_SIN(ia
)->sin_addr
,
1609 sizeof(struct in_addr
));
1610 cp
[IPOPT_OFFSET
] += sizeof(struct in_addr
);
1614 if (ipstealth
&& pass
== 0)
1616 code
= cp
- (u_char
*)ip
;
1617 if (optlen
< 4 || optlen
> 40) {
1618 code
= &cp
[IPOPT_OLEN
] - (u_char
*)ip
;
1621 if ((off
= cp
[IPOPT_OFFSET
]) < 5) {
1622 code
= &cp
[IPOPT_OLEN
] - (u_char
*)ip
;
1625 if (off
> optlen
- (int)sizeof(int32_t)) {
1626 cp
[IPOPT_OFFSET
+ 1] += (1 << 4);
1627 if ((cp
[IPOPT_OFFSET
+ 1] & 0xf0) == 0) {
1628 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1633 off
--; /* 0 origin */
1634 sin
= (struct in_addr
*)(cp
+ off
);
1635 switch (cp
[IPOPT_OFFSET
+ 1] & 0x0f) {
1637 case IPOPT_TS_TSONLY
:
1640 case IPOPT_TS_TSANDADDR
:
1641 if (off
+ sizeof(n_time
) +
1642 sizeof(struct in_addr
) > optlen
) {
1643 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1646 ipaddr
.sin_addr
= dst
;
1647 ia
= (INA
)ifaof_ifpforaddr((SA
)&ipaddr
,
1651 memcpy(sin
, &IA_SIN(ia
)->sin_addr
,
1652 sizeof(struct in_addr
));
1653 cp
[IPOPT_OFFSET
] += sizeof(struct in_addr
);
1654 off
+= sizeof(struct in_addr
);
1657 case IPOPT_TS_PRESPEC
:
1658 if (off
+ sizeof(n_time
) +
1659 sizeof(struct in_addr
) > optlen
) {
1660 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1663 memcpy(&ipaddr
.sin_addr
, sin
,
1664 sizeof(struct in_addr
));
1665 if (ifa_ifwithaddr((SA
)&ipaddr
) == NULL
)
1667 cp
[IPOPT_OFFSET
] += sizeof(struct in_addr
);
1668 off
+= sizeof(struct in_addr
);
1672 code
= &cp
[IPOPT_OFFSET
+ 1] - (u_char
*)ip
;
1676 memcpy(cp
+ off
, &ntime
, sizeof(n_time
));
1677 cp
[IPOPT_OFFSET
] += sizeof(n_time
);
1680 if (forward
&& ipforwarding
) {
1681 ip_forward(m
, TRUE
, next_hop
);
1686 icmp_error(m
, type
, code
, 0, 0);
1687 ipstat
.ips_badoptions
++;
1692 * Given address of next destination (final or next hop),
1693 * return internet address info of interface to be used to get there.
1696 ip_rtaddr(struct in_addr dst
, struct route
*ro
)
1698 struct sockaddr_in
*sin
;
1700 sin
= (struct sockaddr_in
*)&ro
->ro_dst
;
1702 if (ro
->ro_rt
== NULL
|| dst
.s_addr
!= sin
->sin_addr
.s_addr
) {
1703 if (ro
->ro_rt
!= NULL
) {
1707 sin
->sin_family
= AF_INET
;
1708 sin
->sin_len
= sizeof *sin
;
1709 sin
->sin_addr
= dst
;
1710 rtalloc_ign(ro
, RTF_PRCLONING
);
1713 if (ro
->ro_rt
== NULL
)
1716 return (ifatoia(ro
->ro_rt
->rt_ifa
));
1720 * Save incoming source route for use in replies,
1721 * to be picked up later by ip_srcroute if the receiver is interested.
1724 save_rte(u_char
*option
, struct in_addr dst
)
1728 olen
= option
[IPOPT_OLEN
];
1731 kprintf("save_rte: olen %d\n", olen
);
1733 if (olen
> sizeof(ip_srcrt
) - (1 + sizeof(dst
)))
1735 bcopy(option
, ip_srcrt
.srcopt
, olen
);
1736 ip_nhops
= (olen
- IPOPT_OFFSET
- 1) / sizeof(struct in_addr
);
1741 * Retrieve incoming source route for use in replies,
1742 * in the same form used by setsockopt.
1743 * The first hop is placed before the options, will be removed later.
1748 struct in_addr
*p
, *q
;
1753 m
= m_get(MB_DONTWAIT
, MT_HEADER
);
1757 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1759 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1760 m
->m_len
= ip_nhops
* sizeof(struct in_addr
) + sizeof(struct in_addr
) +
1764 kprintf("ip_srcroute: nhops %d mlen %d", ip_nhops
, m
->m_len
);
1768 * First save first hop for return route
1770 p
= &ip_srcrt
.route
[ip_nhops
- 1];
1771 *(mtod(m
, struct in_addr
*)) = *p
--;
1774 kprintf(" hops %x", ntohl(mtod(m
, struct in_addr
*)->s_addr
));
1778 * Copy option fields and padding (nop) to mbuf.
1780 ip_srcrt
.nop
= IPOPT_NOP
;
1781 ip_srcrt
.srcopt
[IPOPT_OFFSET
] = IPOPT_MINOFF
;
1782 memcpy(mtod(m
, caddr_t
) + sizeof(struct in_addr
), &ip_srcrt
.nop
,
1784 q
= (struct in_addr
*)(mtod(m
, caddr_t
) +
1785 sizeof(struct in_addr
) + OPTSIZ
);
1788 * Record return path as an IP source route,
1789 * reversing the path (pointers are now aligned).
1791 while (p
>= ip_srcrt
.route
) {
1794 kprintf(" %x", ntohl(q
->s_addr
));
1799 * Last hop goes to final destination.
1804 kprintf(" %x\n", ntohl(q
->s_addr
));
1810 * Strip out IP options.
1813 ip_stripoptions(struct mbuf
*m
)
1816 struct ip
*ip
= mtod(m
, struct ip
*);
1820 optlen
= (IP_VHL_HL(ip
->ip_vhl
) << 2) - sizeof(struct ip
);
1821 opts
= (caddr_t
)(ip
+ 1);
1822 datalen
= m
->m_len
- (sizeof(struct ip
) + optlen
);
1823 bcopy(opts
+ optlen
, opts
, datalen
);
1825 if (m
->m_flags
& M_PKTHDR
)
1826 m
->m_pkthdr
.len
-= optlen
;
1827 ip
->ip_vhl
= IP_MAKE_VHL(IPVERSION
, sizeof(struct ip
) >> 2);
1830 u_char inetctlerrmap
[PRC_NCMDS
] = {
1832 0, EMSGSIZE
, EHOSTDOWN
, EHOSTUNREACH
,
1833 EHOSTUNREACH
, EHOSTUNREACH
, ECONNREFUSED
, ECONNREFUSED
,
1834 EMSGSIZE
, EHOSTUNREACH
, 0, 0,
1836 ENOPROTOOPT
, ECONNREFUSED
1840 * Forward a packet. If some error occurs return the sender
1841 * an icmp packet. Note we can't always generate a meaningful
1842 * icmp message because icmp doesn't have a large enough repertoire
1843 * of codes and types.
1845 * If not forwarding, just drop the packet. This could be confusing
1846 * if ipforwarding was zero but some routing protocol was advancing
1847 * us as a gateway to somewhere. However, we must let the routing
1848 * protocol deal with that.
1850 * The using_srcrt parameter indicates whether the packet is being forwarded
1851 * via a source route.
1854 ip_forward(struct mbuf
*m
, boolean_t using_srcrt
, struct sockaddr_in
*next_hop
)
1856 struct ip
*ip
= mtod(m
, struct ip
*);
1857 struct sockaddr_in
*ipforward_rtaddr
;
1859 int error
, type
= 0, code
= 0, destmtu
= 0;
1862 struct in_addr pkt_dst
;
1863 struct route
*cache_rt
= &ipforward_rt
[mycpuid
];
1867 * Cache the destination address of the packet; this may be
1868 * changed by use of 'ipfw fwd'.
1870 pkt_dst
= (next_hop
!= NULL
) ? next_hop
->sin_addr
: ip
->ip_dst
;
1874 kprintf("forward: src %x dst %x ttl %x\n",
1875 ip
->ip_src
.s_addr
, pkt_dst
.s_addr
, ip
->ip_ttl
);
1878 if (m
->m_flags
& (M_BCAST
| M_MCAST
) || !in_canforward(pkt_dst
)) {
1879 ipstat
.ips_cantforward
++;
1883 if (!ipstealth
&& ip
->ip_ttl
<= IPTTLDEC
) {
1884 icmp_error(m
, ICMP_TIMXCEED
, ICMP_TIMXCEED_INTRANS
, dest
, 0);
1888 ipforward_rtaddr
= (struct sockaddr_in
*) &cache_rt
->ro_dst
;
1889 if (cache_rt
->ro_rt
== NULL
||
1890 ipforward_rtaddr
->sin_addr
.s_addr
!= pkt_dst
.s_addr
) {
1891 if (cache_rt
->ro_rt
!= NULL
) {
1892 RTFREE(cache_rt
->ro_rt
);
1893 cache_rt
->ro_rt
= NULL
;
1895 ipforward_rtaddr
->sin_family
= AF_INET
;
1896 ipforward_rtaddr
->sin_len
= sizeof(struct sockaddr_in
);
1897 ipforward_rtaddr
->sin_addr
= pkt_dst
;
1898 rtalloc_ign(cache_rt
, RTF_PRCLONING
);
1899 if (cache_rt
->ro_rt
== NULL
) {
1900 icmp_error(m
, ICMP_UNREACH
, ICMP_UNREACH_HOST
, dest
, 0);
1904 rt
= cache_rt
->ro_rt
;
1907 * Save the IP header and at most 8 bytes of the payload,
1908 * in case we need to generate an ICMP message to the src.
1910 * XXX this can be optimized a lot by saving the data in a local
1911 * buffer on the stack (72 bytes at most), and only allocating the
1912 * mbuf if really necessary. The vast majority of the packets
1913 * are forwarded without having to send an ICMP back (either
1914 * because unnecessary, or because rate limited), so we are
1915 * really we are wasting a lot of work here.
1917 * We don't use m_copy() because it might return a reference
1918 * to a shared cluster. Both this function and ip_output()
1919 * assume exclusive access to the IP header in `m', so any
1920 * data in a cluster may change before we reach icmp_error().
1922 MGETHDR(mcopy
, MB_DONTWAIT
, m
->m_type
);
1923 if (mcopy
!= NULL
&& !m_dup_pkthdr(mcopy
, m
, MB_DONTWAIT
)) {
1925 * It's probably ok if the pkthdr dup fails (because
1926 * the deep copy of the tag chain failed), but for now
1927 * be conservative and just discard the copy since
1928 * code below may some day want the tags.
1933 if (mcopy
!= NULL
) {
1934 mcopy
->m_len
= imin((IP_VHL_HL(ip
->ip_vhl
) << 2) + 8,
1936 mcopy
->m_pkthdr
.len
= mcopy
->m_len
;
1937 m_copydata(m
, 0, mcopy
->m_len
, mtod(mcopy
, caddr_t
));
1941 ip
->ip_ttl
-= IPTTLDEC
;
1944 * If forwarding packet using same interface that it came in on,
1945 * perhaps should send a redirect to sender to shortcut a hop.
1946 * Only send redirect if source is sending directly to us,
1947 * and if packet was not source routed (or has any options).
1948 * Also, don't send redirect if forwarding using a default route
1949 * or a route modified by a redirect.
1951 if (rt
->rt_ifp
== m
->m_pkthdr
.rcvif
&&
1952 !(rt
->rt_flags
& (RTF_DYNAMIC
| RTF_MODIFIED
)) &&
1953 satosin(rt_key(rt
))->sin_addr
.s_addr
!= INADDR_ANY
&&
1954 ipsendredirects
&& !using_srcrt
&& next_hop
== NULL
) {
1955 u_long src
= ntohl(ip
->ip_src
.s_addr
);
1956 struct in_ifaddr
*rt_ifa
= (struct in_ifaddr
*)rt
->rt_ifa
;
1958 if (rt_ifa
!= NULL
&&
1959 (src
& rt_ifa
->ia_subnetmask
) == rt_ifa
->ia_subnet
) {
1960 if (rt
->rt_flags
& RTF_GATEWAY
)
1961 dest
= satosin(rt
->rt_gateway
)->sin_addr
.s_addr
;
1963 dest
= pkt_dst
.s_addr
;
1965 * Router requirements says to only send
1968 type
= ICMP_REDIRECT
;
1969 code
= ICMP_REDIRECT_HOST
;
1972 kprintf("redirect (%d) to %x\n", code
, dest
);
1977 error
= ip_output(m
, NULL
, cache_rt
, IP_FORWARDING
, NULL
, NULL
);
1979 ipstat
.ips_forward
++;
1982 ipflow_create(cache_rt
, mcopy
);
1985 return; /* most common case */
1987 ipstat
.ips_redirectsent
++;
1990 ipstat
.ips_cantforward
++;
1997 * Send ICMP message.
2002 case 0: /* forwarded, but need redirect */
2003 /* type, code set above */
2006 case ENETUNREACH
: /* shouldn't happen, checked above */
2011 type
= ICMP_UNREACH
;
2012 code
= ICMP_UNREACH_HOST
;
2016 type
= ICMP_UNREACH
;
2017 code
= ICMP_UNREACH_NEEDFRAG
;
2020 * If the packet is routed over IPsec tunnel, tell the
2021 * originator the tunnel MTU.
2022 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2025 if (cache_rt
->ro_rt
!= NULL
) {
2026 struct secpolicy
*sp
= NULL
;
2031 sp
= ipsec4_getpolicybyaddr(mcopy
,
2037 destmtu
= cache_rt
->ro_rt
->rt_ifp
->if_mtu
;
2039 /* count IPsec header size */
2040 ipsechdr
= ipsec4_hdrsiz(mcopy
,
2045 * find the correct route for outer IPv4
2046 * header, compute tunnel MTU.
2049 if (sp
->req
!= NULL
&& sp
->req
->sav
!= NULL
&&
2050 sp
->req
->sav
->sah
!= NULL
) {
2051 ro
= &sp
->req
->sav
->sah
->sa_route
;
2052 if (ro
->ro_rt
!= NULL
&&
2053 ro
->ro_rt
->rt_ifp
!= NULL
) {
2055 ro
->ro_rt
->rt_ifp
->if_mtu
;
2056 destmtu
-= ipsechdr
;
2065 * If the packet is routed over IPsec tunnel, tell the
2066 * originator the tunnel MTU.
2067 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2070 if (cache_rt
->ro_rt
!= NULL
) {
2071 struct secpolicy
*sp
= NULL
;
2076 sp
= ipsec_getpolicybyaddr(mcopy
,
2082 destmtu
= cache_rt
->ro_rt
->rt_ifp
->if_mtu
;
2084 /* count IPsec header size */
2085 ipsechdr
= ipsec4_hdrsiz(mcopy
,
2090 * find the correct route for outer IPv4
2091 * header, compute tunnel MTU.
2094 if (sp
->req
!= NULL
&&
2095 sp
->req
->sav
!= NULL
&&
2096 sp
->req
->sav
->sah
!= NULL
) {
2097 ro
= &sp
->req
->sav
->sah
->sa_route
;
2098 if (ro
->ro_rt
!= NULL
&&
2099 ro
->ro_rt
->rt_ifp
!= NULL
) {
2101 ro
->ro_rt
->rt_ifp
->if_mtu
;
2102 destmtu
-= ipsechdr
;
2109 #else /* !IPSEC && !FAST_IPSEC */
2110 if (cache_rt
->ro_rt
!= NULL
)
2111 destmtu
= cache_rt
->ro_rt
->rt_ifp
->if_mtu
;
2113 ipstat
.ips_cantfrag
++;
2118 * A router should not generate ICMP_SOURCEQUENCH as
2119 * required in RFC1812 Requirements for IP Version 4 Routers.
2120 * Source quench could be a big problem under DoS attacks,
2121 * or if the underlying interface is rate-limited.
2122 * Those who need source quench packets may re-enable them
2123 * via the net.inet.ip.sendsourcequench sysctl.
2125 if (!ip_sendsourcequench
) {
2129 type
= ICMP_SOURCEQUENCH
;
2134 case EACCES
: /* ipfw denied packet */
2138 icmp_error(mcopy
, type
, code
, dest
, destmtu
);
2142 ip_savecontrol(struct inpcb
*inp
, struct mbuf
**mp
, struct ip
*ip
,
2145 if (inp
->inp_socket
->so_options
& SO_TIMESTAMP
) {
2149 *mp
= sbcreatecontrol((caddr_t
) &tv
, sizeof(tv
),
2150 SCM_TIMESTAMP
, SOL_SOCKET
);
2152 mp
= &(*mp
)->m_next
;
2154 if (inp
->inp_flags
& INP_RECVDSTADDR
) {
2155 *mp
= sbcreatecontrol((caddr_t
) &ip
->ip_dst
,
2156 sizeof(struct in_addr
), IP_RECVDSTADDR
, IPPROTO_IP
);
2158 mp
= &(*mp
)->m_next
;
2160 if (inp
->inp_flags
& INP_RECVTTL
) {
2161 *mp
= sbcreatecontrol((caddr_t
) &ip
->ip_ttl
,
2162 sizeof(u_char
), IP_RECVTTL
, IPPROTO_IP
);
2164 mp
= &(*mp
)->m_next
;
2168 * Moving these out of udp_input() made them even more broken
2169 * than they already were.
2171 /* options were tossed already */
2172 if (inp
->inp_flags
& INP_RECVOPTS
) {
2173 *mp
= sbcreatecontrol((caddr_t
) opts_deleted_above
,
2174 sizeof(struct in_addr
), IP_RECVOPTS
, IPPROTO_IP
);
2176 mp
= &(*mp
)->m_next
;
2178 /* ip_srcroute doesn't do what we want here, need to fix */
2179 if (inp
->inp_flags
& INP_RECVRETOPTS
) {
2180 *mp
= sbcreatecontrol((caddr_t
) ip_srcroute(),
2181 sizeof(struct in_addr
), IP_RECVRETOPTS
, IPPROTO_IP
);
2183 mp
= &(*mp
)->m_next
;
2186 if (inp
->inp_flags
& INP_RECVIF
) {
2189 struct sockaddr_dl sdl
;
2192 struct sockaddr_dl
*sdp
;
2193 struct sockaddr_dl
*sdl2
= &sdlbuf
.sdl
;
2195 if (((ifp
= m
->m_pkthdr
.rcvif
)) &&
2196 ((ifp
->if_index
!= 0) && (ifp
->if_index
<= if_index
))) {
2197 sdp
= IF_LLSOCKADDR(ifp
);
2199 * Change our mind and don't try copy.
2201 if ((sdp
->sdl_family
!= AF_LINK
) ||
2202 (sdp
->sdl_len
> sizeof(sdlbuf
))) {
2205 bcopy(sdp
, sdl2
, sdp
->sdl_len
);
2209 offsetof(struct sockaddr_dl
, sdl_data
[0]);
2210 sdl2
->sdl_family
= AF_LINK
;
2211 sdl2
->sdl_index
= 0;
2212 sdl2
->sdl_nlen
= sdl2
->sdl_alen
= sdl2
->sdl_slen
= 0;
2214 *mp
= sbcreatecontrol((caddr_t
) sdl2
, sdl2
->sdl_len
,
2215 IP_RECVIF
, IPPROTO_IP
);
2217 mp
= &(*mp
)->m_next
;
2222 * XXX these routines are called from the upper part of the kernel.
2224 * They could also be moved to ip_mroute.c, since all the RSVP
2225 * handling is done there already.
2228 ip_rsvp_init(struct socket
*so
)
2230 if (so
->so_type
!= SOCK_RAW
||
2231 so
->so_proto
->pr_protocol
!= IPPROTO_RSVP
)
2234 if (ip_rsvpd
!= NULL
)
2239 * This may seem silly, but we need to be sure we don't over-increment
2240 * the RSVP counter, in case something slips up.
2255 * This may seem silly, but we need to be sure we don't over-decrement
2256 * the RSVP counter, in case something slips up.
2266 rsvp_input(struct mbuf
*m
, ...) /* XXX must fixup manually */
2272 off
= __va_arg(ap
, int);
2273 proto
= __va_arg(ap
, int);
2276 if (rsvp_input_p
) { /* call the real one if loaded */
2277 rsvp_input_p(m
, off
, proto
);
2281 /* Can still get packets with rsvp_on = 0 if there is a local member
2282 * of the group to which the RSVP packet is addressed. But in this
2283 * case we want to throw the packet away.
2291 if (ip_rsvpd
!= NULL
) {
2292 rip_input(m
, off
, proto
);
2295 /* Drop the packet */